Methodology

Exposure and Health Impact

Health impact assessment begins with estimating the population exposure to various air pollutants. We built global concentration maps by combining widely-used baseline maps (namely Larkin et al. for NO2 and Hammer et al. for PM2.5) and ground-measurements collected from various sources:

European Environmental Agency (Europe)
Ministry of Ecology and Environment (China)
US AirNow (US)
DEFRA (UK)
Central Pollution Control Board (CPCB)
Ministry of Environment and Urbanisation (Turkey)
Air4Thai (Thailand)
Environmental Management Bureau (Philippines)
AEROS (Japan)
SAAQIS (South Africa)
OpenAQ (Chile, Taiwan, and Australia)

Ground measurements were taken from the year 2019 and only in cities with measurements available more than 80% of the days that year. That amounted to 3,236 cities with valid NO2 and 2,782 cities with valid PM2.5 measurements. Literature-based basemaps were then updated with these measurements using generalized additive models (GAMs) including a spatial smoothing term. Only statistically significant differences were added to the baselines, and within 0.26 degrees of urban areas (where 95% of the monitoring stations lie).

Finally, a population x exposure distribution was built for each individual country, allowing for fast health impact assessment, yet more accurate than a single population-weighted average exposure. The health impacts are assessed following the methodology of the CREA publication "Quantifying the Economic Costs of Air Pollution from Fossil Fuels".

Following the update of WHO Air Quality Guidelines, which now recognize health harm from NO2 at low concentrations, we have updated the mortality risk function for NO2 based on the findings of Faustini et al. 2014, and including impacts down to 4.5 µg/m3, the lowest concentration level in studies that found increased mortality risk.

NCAP Cities

The average yearly (financial year; FY) PM10 concentration was obtained using the hourly data available from the CPCB website. The Indian financial year starts on 1 April of the previous year to 31 March of the financial year (i.e., FY24 = 1 April 2023 - 31 March 2024). It was first aggregated by city, then daily, and lastly yearly. The yearly average was only calculated when there are at least 245 days (about two-thirds of a year) of observed data, except for the latest financial year. In that case, the minimum data was two-thirds of the number of days passed in that financial year.

For the target of each NCAP city, there are two types of targets. The first one is a predetermined number and the other is a 15% reduction based on the previous FY’s average. In the latter case, the previous FY’s average may not already be available since it includes both automatic and manual measurements. Therefore, the targets for FY2023 and FY2024, as of 1 September 2023, were computed solely from the automatic measurement data which are available.

All Cities

The average monthly or yearly (calendar year) concentrations of three air pollutants; PM10, PM2.5, and ozone, were calculated based on the data available from the CPCB website. The data was first averaged by city, then daily, then yearly. The minimum number of observations required were 20 days for monthly average and 245 days for yearly average.

AQ Trends

Air quality data for each station are obtained from the CPCB website. They are then cleaned from outlier values using Median Absolute Deviation (MAD) with a 70th percentile and a multiplier of 15 before being averaged by date. The city average was then calculated using the mean value for the stations located within each city. The AQ trends shown is a 365 days running average for PM10 and PM2.5. The minimum number of observations for the running average was 290 days (~80%).

Interventions

We tried to separate the effects of weather and air pollution interventions to see how much they influence the air quality in different cities. Based on the result of the algorithm, the contribution of weather and interventions towards the change in air quality of one month compared to the same month